Interpreting a Classical Geometric Proof with Interactive Realizability

Abstract: We show how to extract a monotonic learning algorithm from a classical proof of a geometric statement by interpreting the proof by means of interactive realizability, a realizability sematics for classical logic. The statement is about the existence of a convex angle including a finite collections of points in the real plane and it is related to the existence of a convex hull. We define real numbers as Cauchy sequences of rational numbers, therefore equality and ordering are not decidable. While the proof looks superficially constructive, it employs classical reasoning to handle undecidable comparisons between real numbers, making the underlying algorithm non-effective. The interactive realizability interpretation transforms the non-effective linear algorithm described by the proof into an effective one that uses backtracking to learn from its mistakes. The effective algorithm exhibits a "smart" behavior, performing comparisons only up to the precision required to prove the final statement. This behavior is not explicitly planned but arises from the interactive interpretation of comparisons between Cauchy sequences.